next up previous contents
Next: Robust Estimation Up: Kalman Filtering Technique Previous: Iterated Extended Kalman Filter

Application to Conic Fitting

Let us choose the normalization with A+C=1 (see Sect.4.1). The state vector can now be defined as:

displaymath3428

The measurement vector is: tex2html_wrap_inline3043 . As the conic parameters are the same for all points, we have the following simple system equation:

displaymath3429

and the noise term tex2html_wrap_inline3153 is zero. The observation function is

displaymath3430

In order to apply the extended Kalman filter, we need to compute the derivative of tex2html_wrap_inline3440 with respect to tex2html_wrap_inline3149 and that with respect to tex2html_wrap_inline2849 , which are given by

eqnarray1191



Zhengyou Zhang
Thu Feb 8 11:42:20 MET 1996